markovRing -- ring of joint probability distributions on several discrete random variables

Synopsis

• Usage:
markovRing(d)
• Inputs:
• d, , with positive integer entries $(d_1,\dots ,d_r)$
• Optional inputs:
• Coefficients => ..., default value QQ, optional input to choose the base field in markovRing or gaussianRing
• VariableName => ..., default value p, symbol used for indeterminates in a ring of discrete joint probability distributions
• Outputs:
• a ring, a polynomial ring with $d_1*d_2*\dots *d_r$ variables $p_{i_1,\dots ,i_r}$, with each $i_j$ satisfying $1\leq i_j \leq d_j$.
• Consequences:
• Information about this sequence of integers is placed into the ring, and is used by other functions in this package. Also, at most one ring for each such sequence is created since the ring is cached.

Description

The sequence $d$ represents the number of states of each discrete random variable. This example creates a ring of joint probability distributions on 4 random variables with 2, 3, 4, and 5 states. This ring has a total of 120 indeterminates.

 i1 : d=(2,3,4,5); i2 : R = markovRing d; i3 : numgens R o3 = 120 i4 : R_0, R_1, R_119 --here are some of the variables in the ring o4 = (p , p , p ) 1,1,1,1 1,1,1,2 2,3,4,5 o4 : Sequence

If no coefficient choice is specified, the polynomial ring is created over the rationals.

 i5 : coefficientRing R o5 = QQ o5 : Ring

The optional argument Coefficients allows to change the base field.

 i6 : R2 = markovRing (d,Coefficients=>CC); i7 : coefficientRing R2 o7 = CC 53 o7 : ComplexField

The indeterminates are labeled with the letter ''p'' suggesting probability distributions. However, it is useful to be able to create a new ring where the indeterminates are labeled different (for example, they may represent marginal probabilities). This can be accomplished with the VariableName option.

 i8 : d=(1,2); i9 : markovRing (d,VariableName => q); i10 : gens oo o10 = {q , q } 1,1 1,2 o10 : List

The routines hiddenMap, inverseMarginMap, marginMap, markovMatrices require the ring to be created by this function. The routines conditionalIndependenceIdeal, discreteVanishingIdeal require the ring to be created by this function or the method gaussianRing.

Caveat

As opposed to gaussianRing, this method does not store information about a graph or the names of the random variables. In case these random variables are not numbered $1, 2, \dots, n$, then the methods conditionalIndependenceIdeal and markovMatrices require an additional input in the form of a list of the random variable names. This list must be in the same order as the implicit order used in the sequence $d$. The user is encouraged to read the caveat on the method conditionalIndependenceIdeal regarding probability distributions on discrete random variables that have been labeled arbitrarily.

• conditionalIndependenceIdeal -- ideal of a list of conditional independent statements
• discreteVanishingIdeal -- vanishing ideal of a discrete graphical model
• gaussianRing -- ring of Gaussian correlations on n random variables or a graphical model
• hiddenMap -- linear map between the ring of a model with one hidden variable and the ring of the corresponding fully observed model
• inverseMarginMap -- inverse of the marginMap
• marginMap -- linear map on joint distributions for discrete random variables replacing marginals for indeterminates
• markovMatrices -- matrices whose minors form the ideal of a list of independence statements

Ways to use markovRing :

• "markovRing(Sequence)"

For the programmer

The object markovRing is .